Download Do you two know each other? Transitivity, homophily

Journal of Personality and Social Psychology
2010, Vol. 99, No. 5, 855– 869
© 2010 American Psychological Association
0022-3514/10/$12.00 DOI: 10.1037/a0020961
Do You Two Know Each Other?
Transitivity, Homophily, and the Need for (Network) Closure
Francis J. Flynn
Ray E. Reagans
Stanford University
Massachusetts Institute of Technology
Lucia Guillory
Stanford University
The authors investigate whether need for closure affects how people seek order in judging social
relations. In Study 1, the authors find that people who have a high need for closure (NFC) were more
likely to assume their social contacts were connected to each other (i.e., transitivity) when this was not
the case. In Studies 2 and 3, the authors examine another form of order in network relations—racial
homophily—and find that high-NFC participants were more inclined to believe that 2 individuals from
the same racial category (e.g., African American) were friends than two racially dissimilar individuals.
Furthermore, high-NFC individuals were more likely to make errors when judging a racially mixed group
of people; specifically, they recalled more racial homophily (racially similar people sitting closer
together) than had actually appeared.
Keywords: need for closure, cognitive networks, homophily, race
Our aim in the present research was to make three central
contributions to the literatures on NFC and social networks. First,
we hypothesize that an individual’s NFC affects his or her perception of social networks (i.e., who knows whom). Second, we
introduce the concept of network closure to theory and research on
cognitive closure. Specifically, we propose that NFC (a personality
trait) and network closure (a social structural condition) have a
strong, positive connection. Third, we propose that the association
between NFCand how people judge the presence of a social
relationship will be most pronounced in judgments of racially
similar dyads, thereby adding to a growing stream of research on
the connection between NFCand perceived homogeneity.
People often make mistakes in judging social relations among
their peers (DeSoto, Henley, & London, 1968; Rubin & Zajonc,
1969; von Hecker, 1997; Zajonc & Sherman, 1967 ). For example,
rather than assume that two individuals lack a relationship, a focal
actor will tend to overestimate the extent to which a social relationship exists. Previous research suggests that common personality constructs cannot explain individual differences in perceiving
network relations (Janicik & Larrick, 2005). However, we propose
that some individuals, particularly those who have a high need for
closure (NFC), are prone to commit these errors because it satisfies
their preference for order in understanding social relations.
We advance the idea that cognitive closure and cognitive networks are linked in a way that can account for errors in judging the
existence of interpersonal relations. In particular, we suspect that
individuals with a high NFC will look for order in patterns of
network ties in order to avoid feelings of ambiguity. We focus here
on two forms of order in social relations: transitivity and racial
homophily. Both represent key areas of interest for network researchers and are stable and reliable trends in how network connections emerge (Lin, 2001; Louch, 2000). Indeed, the concept of
closure (connecting two direct contacts) is considered a critical
characteristic of social networks (Burt, 1992), and racial homophily is generally viewed as a defining feature of network connections in the United States (McPherson, Smith-Lovin, & Cook,
2001, p. 420).
NFC
According to Kruglanski (1990), NFC refers to an individual’s
desire for “an answer on a given topic, any answer, as compared to
confusion and ambiguity” (p. 337). Need for closure can be situationally induced (e.g., by an increase in time pressure or ambient
noise), and it can vary in strength across individuals (Webster &
Kruglanski, 1994) and cultures (Mannetti, Pierro, Kruglanski,
Taris, & Bezinovic, 2002). As a personality trait, NFC represents
a stable difference in an individual’s motivation to engage in
elaborate thought processes (Neuberg, Judice, & West, 1997). A
person with a high NFC prefers order and predictability and
experiences discomfort when faced with ambiguity (Van Hiel &
Mervielde, 2003). People who rate low on NFC tend to be less
decisive and more open-minded, which enables them to express
more ideational fluidity and creativity (Chirumbolo, Livi, Mannetti, Pierro, & Kruglanski, 2004).
NFC has been conceptualized as a unidimensional construct, but
both theory and research point to its multidimensional quality
(Kruglanski et al., 1997; Neuberg et al., 1997). Empirical analyses
Francis J. Flynn and Lucia Guillory, Graduate School of Business,
Stanford University; Ray E. Regans, Sloan School of Management, Massachusetts Institute of Technology.
Correspondence concerning this article should be addressed to Francis J.
Flynn, Department of Psychology, Stanford University, 518 Memorial
Way, Stanford, CA 94305. E-mail: [email protected]
855
FLYNN, REAGANS, AND GUILLORY
856
of measures of NFC reveal two underlying factors that are related,
but not completely overlapping. One factor refers to the desire for
decisiveness, which is the motivation to attain closure as soon as
possible, and the other refers to the need for simple structure,
which is the motivation to maintain closure for as long as possible.
Kruglanski and Webster (1996) described these two dimensions of
NFC as “seizing” (a tendency to value decisiveness in oneself and
others) and “freezing” (a tendency to maintain intuitions, attitudes,
and social rules with some degree of permanence). Seizing reflects
a sense of urgency, in which delays are viewed as bothersome and
inefficient. Freezing reflects a preference for permanence, in
which knowledge is safeguarded and assumptions are held fast.
Perceiving Network Closure
Social networks depict the pattern of interpersonal relations
among a set of individuals (Burt, 1992; Granovetter, 1973). Such
patterns can help explain many critical outcomes, ranging from
conferrals of power and status (e.g., Brass & Burkhardt, 1993) to
intraorganizational job mobility and performance (e.g., Podolny &
Baron, 1997). One of the most heavily studied patterns in social
networks is network closure. From the perspective of a focal
individual (A) who is connected to two peers (B and C), cases in
which the peers are not directly connected (B does not know C) are
referred to as “structural holes” by network researchers, and cases
in which the two peers are connected (A, B, and C all know one
another) are referred to as “network closure.” Structural holes have
been shown to be advantageous in cases in which being a broker
can afford strategic gains (Burt, 1992), and network closure has
been shown to be advantageous in cases in which trust is essential
for strong performance (Coleman, 1988, 1990). The most advantageous network positions combine both features (Reagans &
Zuckerman, 2008).
Although the study of social networks has long been of interest
to sociologists, the perception of social networks (referred to as
“cognitive networks”) has recently become a keen interest of
social psychologists (e.g., Flynn, Reagans, Amanatullah, & Ames,
2006; Janicik & Larrick, 2005; Kilduff, Crossland, Tsai, & Krackhardt, 2008; Krackhardt & Kilduff, 1999). Can people judge “Who
is friends with whom?” and “Who influences whom?” in their
social groups? As Janicik and Larrick (2005) argue,
accurately learning the pattern of interpersonal relations is often
critical to predicting future behavior and acting appropriately in
important social contexts, such as forming a successful team, inviting
the right people to a gathering, or finding critical allies in a work
organization (p. 348).
Although people may clearly be motivated to accurately judge
their own and others’ network ties, such accuracy may be difficult
to obtain (see Ibarra, Kilduff, & Tsai, 2005). People are generally
less aware of others’ patterns of interaction, or their feelings of
friendship, relative to their own. As a result, we must often judge
the existence of social relations by relying on cues, hearsay, or our
own assumptions. Armed with such indirect sources of information, our judgments may be invalid and full of systematic errors.
Indeed, in a recent study, Kilduff et al. (2008) found that most
people tend to assume their social networks are characterized by
higher levels of network closure (i.e., people arranged in dense
clusters) than is actually the case.
The notion that people make mistakes in judging the existence
of a social tie is well established (Janicik & Larrick, 2005). Early
psychological studies suggested that individuals were prone to
commit numerous errors when learning the structure of ties in
social networks (see De Soto, 1960; Zajonc & Burnstein, 1965a,
1965b; Zajonc & Sherman, 1967). In particular, individuals more
often misperceive social ties by assuming a relationship exists
when it does not rather than assuming a relationship does not exist
when in fact it does (e.g., De Soto et al., 1968; Freeman, 1992;
Goldstone, 2000; Picek, Sherman, & Shiffrin, 1975). This tendency is presumed to characterize all individuals, but we wonder
whether it characterizes some more than others. In particular, are
those individuals who prefer closure more likely to see it in their
social world?
The Link Between NFC and Network Closure
Research on the factors that lead to closure in cognitive networks (i.e., assuming two people have a relationship) has tended to
focus on situational conditions that increase the perception of
closure (e.g., Rubin & Zajonc, 1969; von Hecker, 1997). In contrast, individual differences in perceiving network relations have
only recently begun to receive attention. Janicik and Larrick
(2005) found that people who were more exposed to “missing”
relations could more easily recognize and learn other missing
relations. The authors did not find much evidence that a specific
set of personality traits, including agreeableness, Machiavellianism, and self-monitoring, could influence an individual’s ability to
accurately recall incomplete networks. However, Flynn et al.
(2006) found evidence that high self-monitors were better judges
of others’ relations, in general (though the researchers did not
distinguish between judgments of present and absent relationships).
In judging social relations, people are inclined to assume that
social relations are present rather than absent (De Soto et al., 1968;
Janicik & Larrick, 2005). One special case of this “presence versus
absence” heuristic is the principle of transitivity. According to
Janicik and Larrick,
if a transitive relation exists among three members of a set, such as X,
Y, and Z, then the existence of a relation between X and Y (X r Y) and
between Y and Z (Y r Z) implies that it will also exist between X and
Z (X r Z) (p. 348).
Transitivity is an intuitively appealing rule that can be applied
quickly and easily in attempting to understand a set of social
relations. It may be particularly appealing to individuals with a
high NFC who seek clarity in perceiving the social relations that
surround them.
We suggest that the personality construct of NFC may influence the perception of network closure because it leads people
to overuse the rule of transitivity. Such an effect could be
driven by multiple factors. For example, a critical cause of
cognitive closure is the strong desire for decisiveness in perceiving and understanding the social world (Neuberg et al.,
1997). Such a keen desire for decisiveness can begin with
seizing (prompting people to apply simple heuristics to a wide
range of complex problems) and eventually develop into freezing, or maintaining a firm belief in the value of such rules, even
in the presence of conflicting information (Jost, Glaser,
DO YOU TWO KNOW EACH OTHER?
Kruglanski, & Sulloway, 2003; Tetlock, 2005). This “freezing”
dynamic is often referred to as specific closure, which is the
desire for a specific answer, one that creates or maintains a clear
structure. In the context of judging social relations, transitivity
would provide a specific answer to a complex problem.
Another factor that may lead individuals with a high NFC to
overuse the rule of transitivity is their preference for stability in
distributed knowledge across individuals. Networks with more
structural holes are characterized by a high level of diversity (in
terms of attitudes, perspectives, and access to resources). Conversely, networks with fewer structural holes (i.e., more closure)
tend to be characterized by greater redundancy and consensus (see
Burt, 1992, for evidence). As Kruglanski and Webster have argued
(e.g., Kruglanski & Webster, 1996; Webster & Kruglanski, 1994),
individuals with a high NFC crave shared reality. Such a sense of
shared reality, or social consensus, might only be obtained if
individual members of their social system had relations with each
other and were capable of forming consensus (i.e., the network had
a high level of closure).
857
is a strong feature of network connections in the United States
(e.g., McPherson et al., 2001, p. 420),1 individuals with a high
NFC may seize on similar racial categories as being diagnostic of
some personal affiliation.
There may also be a freezing dynamic that links NFC to
judgments of racial homophily. Individuals with a high NFC
tend to perceive collections of individuals as entitative groups
(Kruglanski et al., 2002). This allows the application of knowledge across individual cases rather than having to form new
knowledge with every new individual, or every new pair of
individuals. In this sense, a high-NFC individual can avoid
epistemic instability in judging social relations by imposing a
sense of order on their understanding of the social world (i.e.,
racial homophily is the rule rather than the exception). Consequently, when people are asked to judge the existence of racially similar and racially dissimilar social relations, those high
in NFC may be more likely to assume social ties exist among
racially similar pairs than racially dissimilar pairs (i.e., presumed racial homophily), whereas those low in NFC will distinguish less between these two cases.
Presumed Racial Homophily and NFC
Overview of Studies and Predictions
The principle of homophily—the tendency for individuals to
associate and bond with similar others— has received rich support
in research by social psychologists, sociologists, and anthropologists. Similar people are physically attracted to one another (Buss
& Barnes, 1986; Watson et al., 2004), tend to communicate more
often (Newcomb, 1961), and are even likely to show mutual
preference in hiring decisions (Sacco, Scheu, Ryan, & Schmitt,
2003). The phenomenon of homophily is well understood by most
individuals, even those who lack sophisticated training and education about human relations. But, are some people less accurate in
detecting homophily? Specifically, do those individuals with a
high NFC overestimate the extent to which homophily characterizes social relations, particularly relationships among members of
the same racial group?
NFC has been linked to racial judgments before. Studies of
authoritarianism, which has long been connected to racial prejudice, show a moderately strong association with NFC (Webster &
Kruglanski, 1994). In addition, Shah, Kruglanski, and Thompson
(1998) found that having a high NFC strengthened ingroup/
outgroup bias: simultaneously increasing ingroup favoritism and
outgroup derogation. A series of studies conducted by Kruglanski,
Shah, Pierro, and Mannetti (2002) went one step further, arguing
that individuals with a high NFC liked ingroups and outgroups
more to the extent that the groups were composed of highly
homogeneous group members. That is, having a high NFC predisposed individual perceivers to evaluate homophilous affiliations
more favorably than heterophilous affiliations.
People with a high NFC clearly prefer homophily, but could it
be that the same people clearly perceive homophily, even in
situations in which it does not exist? When presented with a
question about the existence of a social relationship, such as “Are
John and Mark friends?”, an individual’s NFC may prompt a
narrow information search in an effort to identify the correct
response as soon as possible (Van Hiel & Mervielde, 2003). In
these judgments, one of the most salient pieces of information
available to the focal individual is whether two individuals are
members of the same racial category. Given that racial homophily
We make two central predictions that relate to two powerful
features of social network structure. First, we posit that individuals
with a high NFC will perceive more evidence of network closure
than those individuals with a low NFC. That is, people with a high
NFC will be more likely to assume that two people they know also
know each other when in fact they do not. Second, we expect to
find strong evidence of a “presumed racial homophily” effect in
judgments made by high-NFC individuals of others’ interpersonal
relations.
We tested these predictions in three studies. In Study 1, we
measured people’s perceptions of interpersonal relations among
members of their social group to determine whether NFC led to
greater perceptions of network closure (via the principle of transitivity). In Study 2, we asked people to sketch diagrams of social
relations among a group of unfamiliar targets to test whether those
with a high need for closure were more likely to draw direct ties
among racially similar people rather than racially dissimilar people. In a third study, we asked participants to review and then
recall pictures of people gathered in small groups and investigated
whether participants who had a high NFC were more likely to
erroneously recall racial homophily in each picture.
Study 1
Method
Participants. Fifty-three students enrolled in a 1-year Master’s of Business Administration (MBA) program at a West Coast
university participated in this study. During their time in the
1
We considered whether our effects could be replicated using gender
rather than race as the demographic variable of interest. Our suspicion is
that the effect of gender will not be as pronounced, primarily because
individuals both experience and observe more gender intermixing
(McPherson et al., 2001). Thus, a “gender homophily” heuristic may be
less likely to develop than a “racial homophily” heuristic.
858
FLYNN, REAGANS, AND GUILLORY
program, students were required to take courses with the same
group of fellow students. Of the participants, 75% were men, 42%
were White, 34% were Asian, 20% were Hispanic, and 2% were
Black (4% did not report race). The composition of the group was
stable (no change in membership over the course of the year).
Furthermore, the participants had ample opportunity to observe
interactions among their colleagues because they spent a considerable amount of time with each other, as a group, in and outside
of the classroom.
Procedure. Participants were asked to complete a questionnaire that assessed their relationships with classmates, as well as
their perceptions of relationships among classmates. The questionnaire was administered 6 weeks into the students’ classroom
experience. The design of the questionnaire followed closely on
that originally developed by Krackhardt (1987) and later adapted
by Flynn et al. (2006). The first section presented a complete list
of students in the program. Using this list, participants were asked
to indicate “Whom would you go to for help or for advice if you
had a question or a problem? Such help or advice might include
assistance on a course assignment, copies of notes from classes
you may have missed, career consultations, or other things.” Piloting of the survey suggested that these kinds of behaviors were
both typical and meaningful.
On a separate page, participants were presented with the same list
of individuals, but in this case they were asked to indicate who might
come to them for help or advice. We measured both sides of each
dyadic relation because people can have different impressions of the
same relationship (Krackhardt, 1987; Laumann & Knoke, 1987).
In addition to reporting their own ties, participants were also
asked to describe ties among other members of their class (see
Krackhardt, 1987). Given our classroom size, asking participants
to rate every possible dyadic relation would have been timeconsuming and could have affected the quality of their responses;
instead, we randomly selected five classmates for each participant
and asked him or her to describe the advice/helping relations for
each of these five individuals. Specifically, participants were presented with a customized grid, which included the names of each
individual in the class in the rows and the five randomly assigned
colleagues in the columns. Following others (e.g., Flynn et al.,
2006; Krackhardt, 1987), each participant was asked to indicate
which of their classmates listed in each of the columns would go
to those listed in the rows for help or advice.
Participants were also asked to complete the 47-item Need
for Closure Scale developed and validated by Webster and
Kruglanski (1994). Items were rated on a 6-point scale ranging
from 1 (Strongly disagree) to 6 (Strongly agree) (M ⫽ 152.13,
SD ⫽ 18.308). Sample items include “I think that having clear
rules and order at work is essential for success”; “I dislike
questions which could be answered in several different ways”;
and “I enjoy the uncertainty of going into a new situation
without knowing what might happen” (reverse scored). To
ensure the validity of responses, the Need for Closure Scale
includes five items (e.g., “I have never hurt another person’s
feelings” and “I believe that one should never engage in leisure
activities”) that one would expect to receive low ratings of
agreement. Following the guidelines outlined by Webster and
Kruglanski, we summed a “lie score” for these five items and
removed individuals from the sample who received a score of
15 or higher (n ⫽ 3). For the remaining individuals, we summed
the other 42 items (including reverse-scored items) to create an
overall NFC score. The overall reliability (coefficient alpha) for
this scale was .86.
The response rate for the questionnaire, which took about 20
min to complete, was 96%.
Results
To test our prediction about the link between NFC and transitivity (i.e., perceived network closure), we analyzed participants’
reports of the relationships that exist among their colleagues. For
any relationship between two colleagues, our dependent variable
(“connection in cognitive network”) is binary: The value is 1 if the
respondent believed two of his or her colleagues were connected
by a helping or advice relationship, and the value is 0 if he or she
believed these two colleagues were not connected by such a
relationship. We focus our analysis on cases in which the two
individuals being rated were not connected (according to their own
reports). Thus, any factor that affected the dependent variable
reflected an individual’s ability to recognize structural holes or to
accurately describe missing network connections.
To analyze these data, we used a two-way random effects model
(see Rabe-Hesketh & Skrondal, 2005) rather than other multi-evel
models, like hierarchical linear modeling (HLM). HLM is appropriate when lower order units (e.g., students) are nested within
higher order units (e.g., classrooms), but that is not the case here.
In this case, each respondent is asked to evaluate relationships
among a subset of his or her colleagues, and multiple respondents are
asked to evaluate the same relationship. If each respondent was asked
to evaluate a distinct set of relationships, our relationships would be
nested, but instead our respondents and relationships are “crossed.” A
two-way random effects model is better suited to analyze these
crossed data (Rabe-Hesketh & Skrondal, 2005).
Two other advantages of the two-way random effects model
stand out. First, we have multiple observations for each individual
respondent, and this kind of clustering violates the independence
assumption in regression analysis. The random intercepts for individual respondents allow us to adjust our standard errors for
clustering of this kind. Second, unmeasured features of a specific
pair of classmates could have affected the judgment of their
relationship. In a two-way random effects model, the random
intercepts for classmates allow us to control for unmeasured features of a specific pair of classmates that could have affected how
their relationship was perceived.
We are interested in the tendency for individuals with a high
NFC to assume network connections are transitive, or characterized by closure. To examine this issue, we first created a
categorical variable describing the relationship from the focal
individual to specific pairs of classmates. The focal individual
could have been disconnected from both classmates, connected
to one but not the other, or connected to both. We then created
two categorical variables to measure the strength of the relationship from the respondent to a specific pair of classmates.
Connected to one was set equal to 1 if the focal individual was
connected to only one colleague but remained equal to 0 otherwise. Connected to both was set equal to 1 if the focal
individual was connected to both colleagues but remained equal
to 0 otherwise.
DO YOU TWO KNOW EACH OTHER?
Our dependent variable (connection in cognitive network)
was regressed on the two tie strength variables, the NFC variable, and the interaction between the NFC and two tie strength
variables. We mean centered the NFC variable before multiplying it with each tie strength variable to create the relevant
interaction terms. The coefficients for each variable in the
regression are reported in Table 1. As shown in the table, the
main effects for the tie strength variables are not significantly
different from zero. However, the interactions are significant
and indicate that if a high-NFC individual was connected to one
(␤ ⫽ .030, p ⬍ .05) or both (␤ ⫽ .103, p ⬍ .05) classmates, he
or she was more likely to report the classmates were connected
to each other, whereas low-NFC individuals were less likely to
report the classmates were connected to each other. Given that
the classmates were, in fact, not connected, low-NFC individuals were more likely to describe these relationships accurately.
The empirical results are illustrated in Figure 1. The vertical
axis is the predicted probability the respondent (i.e., A) reports
a tie between two classmates (i.e., B and C) in the cognitive
network. The horizontal axis is NFC. When A was connected to
both colleagues, an increase in NFC was associated with an
increased tendency to report a tie between B and C. Thus, as an
individual’s NFC increased, he or she was more likely to
assume the triad involving B and C was transitive or characterized by closure. We observed a similar pattern when the
respondent was connected to only one colleague, but the increase was not as dramatic.
Finally, when A was not connected to either colleague, an
increase in NFC reduced the likelihood of reporting a tie
between B and C (instances in which A, B, and C are disconnected often are referred to as vacuous transitivity; see Holland
& Leinhardt, 1971, p. 123). The fact that NFC increased the
accuracy of network perception when a triad was vacuous but
reduced accuracy when a triad was imbalanced might indicate
that consensus is a viable explanation for the link between NFC
and transitivity. Shared reality cannot be provided by the absence of ties, and so the tendency for high-NFC individuals to
accurately describe vacuous triads suggests that high-NFC individuals are more likely to exaggerate the presence of relations
rather than their absence.
Noting that previous research using the Need for Closure
Scale has occasionally found different effects for the subscale
of Decisiveness, we separated the NFC measure into two separate measures—the Decisiveness subscale (␣ ⫽ .64) and the
remainder of the scale (i.e., Need for Simple Structure, ␣ ⫽
.87). Looking at these subscales individually, in separate regressions, we do not find any significant interaction effect
between these measures and the tie strength variables on transitivity.
Overall, the empirical results indicate that high-NFC people
have a clear preference for transitivity (in particular, the existence of a relationship) in perceiving social networks. The
pattern of findings in our empirical analysis is consistent with
our hypothesis that individuals with a high NFC would be more
likely to erroneously assume their relationships with others
were transitive (i.e., they are more likely to perceive closure in
their own social relations even when it does not exist).
859
Discussion
In Study 1, we found that individuals who had a high NFC were
more likely to make errors that reflected transitivity, or closure, in
their cognitive networks. In other words, high-NFC individuals
were more prone to assume that two direct contacts in their social
network knew each other when in fact this was not the case
(according to the other parties’ reporting of their own network
ties). Put differently, high-NFC individuals appear to be less able
than low-NFC individuals to recognize structural holes in their
social networks (cf. Janicik & Larrick, 2005).
In the next study, we move from an actual social network to an
artificial network that is constructed by the participant. With these
data, we aim to test our second prediction that high-NFC people
may be especially inclined to assume the existence of a relationship when asked to evaluate racially similar ties, compared with
racially dissimilar ties. Unfortunately, we were unable to test this
idea using the data from Study 1, partly because we did not have
sufficient diversity (there was only one African American student)
and partly because many of the students could not be reliably
categorized as racial minorities by a pair of independent raters
(e.g., many of the Hispanic students were mislabeled as “Caucasian”). Thus, in our second study we turned to a laboratory setting
in which we could control the number of connections in the
network, the ratio of racially similar and dissimilar relationships,
and whether the race of each individual in the network could be
easily recognized.
Study 2
In Study 2, we tested our assumption that people with a high
NFC would show a stronger preference for racially homophilous ties than would individuals with a low NFC by asking
participants to draw network diagrams using photographs of
individuals they had never seen before. We expected that people
with a high NFC would draw a higher percentage of racially
homophilous connections (e.g., two African American photos).
Furthermore, we expected that people with a high NFC would
locate these racially homophilous individuals closer together in
a network diagram (i.e., draw lines of a shorter distance to
connect them).
Method
Participants. Forty-nine students at a West Coast university
participated in this study in exchange for a $10 gift card from a
popular online retailer. Two participants were removed from the
data set for having lie scores that exceeded 15 on the Need for
Closure Scale. Our sample consisted of 20 men (43%) and 27
women (57%); 13 of the participants were White, five were Black,
22 were Asian, four were Hispanic, and three reported their race as
“other.”
Procedure. Participants were asked to draw a diagram of the
relationships that existed between 16 people described to them
only as “fellow students.” At the beginning of the session, participants were given a set of 16 2 in. ⫻ 2 in. color photographs, in
which an individual was shown, from the shoulders up, smiling
directly into the camera. The individuals depicted in the photographs included eight White students (four male and four female),
⫺643.58
36.98
4
8,735
⫺1.125ⴱ (.261)
⫺.004ⴱ (.001)
⫺3.806ⴱ (2.174)
⫺.008 (.014)
⫺.257 (.262)
⫺.732 (.810)
.030ⴱ (.014)
.103ⴱ (.051)
⫺2812.05
181.15
3
5,640
1.346ⴱ (.105)
.010ⴱ (.002)
Connection in
network diagram
Study 1: Connection
in cognitive network
1.107ⴱ (.104)
1.631ⴱ (.191)
2.063ⴱ (.192)
.008ⴱ (.002)
.013ⴱ (.004)
.014ⴱ (.004)
⫺2799.89
246.82
7
5,640
⫺1.112ⴱ (.258)
⫺.004ⴱ (.001)
Connection in
network diagram
⫺20454.56
169.04
3
5,640
4.420ⴱ (.392)
.060ⴱ (.009)
18.280ⴱ (4.449)
.015 (.028)
Closeness in
network diagram
Study 2
Note. Table includes unstandardized regression coefficients with standard error of coefficient in parentheses. NFC ⫽ need for closure.
†
p ⬍ .10. ⴱp ⬍ .05.
Constant
NFC
Connected to one
Connected to both
NFC ⫻ Connected to One
NFC ⫻ Connected to Both
Same race
NFC ⫻ Same Race
Both White
Both Asian American
Both African American
NFC ⫻ Both White
NFC ⫻ Both Asian American
NFC ⫻ Both African American
Log likelihood
Wald chi-square
Degrees of freedom
Number of observations
Predictor variable
Table 1
Summary of Regression Analyses for Studies 1, 2, and 3
3.684ⴱ (.424)
5.865ⴱ (.818)
6.408ⴱ (.818)
.051ⴱ (.010)
.099ⴱ (.020)
.058ⴱ (.020)
⫺20449.86
200.21
7
5,640
18.280ⴱ (4.449)
.015 (.028)
Closeness in
network diagram
⫺33107.32
6.42
3
30,870
⫺.008ⴱ (.008)
.0006ⴱ (.0002)
1.844ⴱ (.075)
⫺.0001 (.0004)
Closeness at
cafeteria table
⫺.009 (.008)
⫺.0137 (.0275)
.0213 (.0275)
.0005† (.0002)
.0018† (.0009)
.0019ⴱ (.0009)
⫺33122.17
11.55
7
30,870
1.844ⴱ (.075)
⫺.0001 (.0004)
Closeness at
cafeteria table
Study 3
860
FLYNN, REAGANS, AND GUILLORY
DO YOU TWO KNOW EACH OTHER?
Figure 1.
861
The positive relationship between need for closure and transitivity in a cognitive network.
four African American students (two male and two female), and
four Asian American students (two male and two female).2 Before
proceeding, the experimenter asked participants whether they
knew any of the individuals shown in the photographs. Each
participant confirmed that they did not.
To ensure the participants could easily recognize the race of the
individuals depicted in the photograph, an independent sample of
50 individuals was asked to review the photos and guess whether
the person shown was (a) White, (b) African American, or (c)
Asian American. Out of 800 responses, only eight were incorrect,
which assured us that the racial categories were clear. Nevertheless, the individuals depicted in the photographs may come across
as “atypical” members of their racial category (see Eberhardt,
Davies, Purdie-Vaughns, & Johnson, 2006). To ensure these individuals were not atypical, members of the independent sample
were asked to rate the typicality of each photo (after they had
sorted the photographs into racial categories). When asked to
evaluate White photos, raters were asked “How much does this
person look like a typical White person?” and instructed to provide
their responses using an 8-point scale ranging from 1 (not typical
at all) to 7 (very typical) (the wording was adjusted for African
American and Asian American photos). The mean typicality scores
for each photo ranged from 4.98 to 6.55, all well above the
midpoint of the scale. More importantly, the typicality ratings did
not differ significantly across racial categories ( p ⫽ .99).
The experimenter explained to the participants that 30 actual
relationships existed among the 16 individuals shown in the photographs (in fact, these individuals were unfamiliar with one another). The participant’s task was to guess “who knows whom”
and to indicate each of these 30 relationships by drawing a line
from one picture to another. Participants were given a 22 in. ⫻ 28
in. poster board and some adhesive tape to construct a network
diagram using the 16 photos. Participants had 15 min to complete
this task (pilot testing confirmed that this was ample time). To
motivate participants to give us their honest guesses of “who
knows whom,” the additional incentive of an iPod Nano was
offered for the participant who was most “accurate” in predicting
the actual network of relationships. One randomly selected participant received the iPod Nano after the study was completed.
Each participant completed their diagrams in the time allotted.
After the participants had finished, the experimenter collected the
diagram materials and replaced them with the Need for Closure
2
We wanted to have at least four people from each minority racial
category (Asian and Black) to minimize the possibility of an idiosyncratic
photo driving the effect (thus two men and two women for both the Asian
and Black category). In addition, we did not wish to go beyond this number
because we did not want participants to think that the study was about race
and that the distribution seemed notably atypical.
862
FLYNN, REAGANS, AND GUILLORY
Scale (Webster & Kruglanski, 1994) described in Study 1. Again,
sample items from the Need for Closure Scale include “I don’t like
situations that are uncertain” and “I enjoy having a clear and
structured mode of life.” Participants were instructed to rate the
extent to which they agreed with each item on a 6-point scale
ranging from 1 (strongly disagree) to 6 (strongly agree). Following Webster and Kruglanski (1994), responses to the scale’s 47
items (excluding the “lie” items) were summed to create an overall
measure of each participant’s need for closure (M ⫽ 151.17, SD ⫽
26.61; ␣ ⫽ .63).3
Results
Given that relationships between photo pairs (i.e., dyadic relationships) were the unit of analysis, we had multiple observations
for each individual participant in our sample. Again, this kind of
clustering violates the independence assumption in regression
analysis. Moreover, unmeasured features of a particular photo pair
could have affected how their relationship was perceived. As in the
previous study, we address these methodological concerns by
estimating a two-way random effects model. The random intercepts for individual respondents allow us to adjust our standard
errors for clustering, and the random intercepts for photo pairs
allow us to control for unmeasured features of a specific pair that
could have affected how their relationship was perceived.
We hypothesized that individuals with a high NFC would predict the existence of more racially homophilous ties than would
individuals with a low NFC. Our dependent variable (connection
in network diagram) was binary and set equal to 1 if the respondent
indicated that the individuals in two photos were connected by
drawing a line between them. The variable was assigned a value of
0 otherwise. Using logistic regression, we regressed our dependent
variable on a categorical variable that was set equal to 1 if the
members of the photo pair were racially similar (both African
American, both Asian American, or both White) and 0 if they were
racially dissimilar, the participant’s NFC score, and an interaction
term (the racial similarity variable multiplied by the participant’s
mean-centered NFC score).
Consistent with our prediction, as a respondent’s NFC increased, he or she was more likely to assume that racially similar
individuals were connected (see Table 1, Column 2; ␤ ⫽ .010, p ⬍
.05). It is possible that our racial similarity effect reflected a
specific kind of racial similarity (i.e., both Asian), rather than a
consistent effect across categories, particularly because there were
more pairs of White photos than African American or Asian
American photos. However, when analyzed separately (see Table
1, Column 3), the effect of NFC was positive and significant for all
three categories of racial similarity: African American (␤ ⫽ .014,
p ⬍ .05) pairings, Asian American (␤ ⫽ .013, p ⬍ .05) pairings,
and White pairings (␤ ⫽ .008, p ⬍ .05).
We expected that participants with a high NFC would be prone
to not only connect people of the same race but also position these
individuals more closely together in terms of physical proximity.
To test this idea, we calculated the distance between every pair of
images in the diagram (N ⫽ 120 pairs). However, noting that some
participants drew much larger network diagrams (photos widely
spaced apart) than did others, we adjusted this measure by subtracting the distance between every pair of photos from the maximum distance between any two pairs of photos in an individual’s
network diagram. Thus, for each of the 120 photo pairs, our
measure of proximity captured the difference between the greatest
distance in a diagram and the distance between the focal pair of
images. Larger values indicate that the photos were placed closer
together.
Our measure of proximity (“closeness in network diagram”) was
regressed on a categorical variable that equaled 1 if the members
of the photo pair were racially similar (0 if they were not), or on
the basis of the participants’ NFC scores and an interaction between the two variables. Again, our NFC variable was mean
centered before calculating the interaction term. We found that
participants with a high NFC were more likely to put racially
similar photos closer together than were people with a low NFC
(see Table 1, Column 4; ␤ ⫽ .060, p ⬍ .05). Again, this effect was
positive and significant (see Table 1, Column 5) for Asian American pairs (␤ ⫽ .099, p ⬍ .05), African American pairs (␤ ⫽ .058,
p ⬍ .05), and White pairs (␤ ⫽ .051, p ⬍ .05). Thus, in addition
to assuming that racially similar individuals were more likely to be
connected to each other, participants with a high NFC were more
likely to cluster racially similar photos in the same geographic
space.4
The empirical results from Study 2 are illustrated in Figures 2
and 3. The vertical axis in Figure 2 is the predicted probability that
the respondent connected a network pair in his or her diagram, and
the vertical axis in Figure 3 is the predicted proximity of a network
pair in the respondent’s network diagram. The horizontal axis in
each case is the respondent’s NFC. Two lines are illustrated in
each figure, one for when the individuals in the network pair were
racially similar (i.e., same race) and another for when they were
racially dissimilar (i.e., different race). The results shown in the
two figures indicate that as an individual’s NFC increased, he or
she was more likely to place racially similar photos closer together
and was more likely to connect racially similar individuals in a
network diagram. Thus, the results support our prediction that
high-NFC individuals were more likely than low-NFC people to
assume racial homophily, in terms of racially similar individuals
being connected to each other and being in close proximity.
Supplementary analyses. Given that the subscale of Decisiveness tends to have different effects from the rest of the Need
for Closure Scale, we conducted separate analyses for decisiveness
(␣ ⫽ .46) and the remainder of the scale (␣ ⫽ .57) to determine
whether one may have been relatively more or less powerful than
the other in explaining these results. As it turns out, the separate
measure of decisiveness had no significant impact, in terms of
moderating the effect of racial similarity on observing a tie in the
network diagram (␤ ⫽ .039, p ⫽ .743) or proximity (␤ ⫽ .368,
p ⫽ .416). However, the rest of the scale, when substituted for the
3
Given this lower alpha, we decided to contact the 49 participants
several weeks later and asked them to complete the NFC scale again. We
were able to gather responses from 45 of the 49 participants, and we found
that, using this follow-up measure (which had an alpha of .88, and a Time
1–Time 2 correlation of. 43, p ⬍ .01), our results from Study 2 replicated.
4
It is possible that our results reflect a tendency for members of one
racial group to assume that members of another racial group are connected
to each other but that such assumptions do not generalize across racial
categories. However, our results hold (here and in Study 3) even when
accounting for these additional control variables.
DO YOU TWO KNOW EACH OTHER?
Figure 2.
863
The positive relationship between need for closure and presumed racial homophily.
overall NFC measure in the interaction term (NFC multiplied by
the general same-race variable), remains a strong predictor of
observing a tie in the network diagram (␤ ⫽ .467, p ⬍ .001) and
proximity (␤ ⫽ 2.42, p ⬍ .05). Albeit far from conclusive, this
result might suggest that decisiveness is a relatively weaker driver
of the perceived racial homophily effect.
with one another. To test whether high-NFC individuals are more
likely to make errors in reporting the presence of racially homophilous ties, we focus on a recall task in Study 3— one in which
accuracy in participants’ reporting can be carefully tracked.
Discussion
In our third study, we turn to a common paradigm in the
judgment of racially similar ties—a cafeteria table. At many college campuses across the United States, students report that cafeteria seating reflects strict racial groupings (e.g., the African American students sit at one table). Although seating is not assigned, the
tendencies toward racial homophily seem overwhelmingly clear to
some observers. In this study, we examined whether such assumptions of racial homophily are exaggerated in the minds of observers, particularly those observers who are high in NFC.
In Study 2, we moved from the field to the lab, asking participants to construct a diagram of a social network using a set of
photographs depicting unfamiliar people and a limited number of
social ties. Forced to predict “who was friends with whom,”
participants were more likely to assume the existence of ties
between racially similar people rather than racially dissimilar
people. More importantly, we found that this effect was stronger
for individuals who had a high NFC relative to those individuals
with a low NFC. In other words, high-NFC individuals were more
prone to assume that two strangers were friends when they were
racially similar to each other.
Although this is clear evidence of presumed racial homophily,
such thinking is not necessarily erroneous. Homophily does exist,
and it may be the case that racial homophily would appear among
this set of individuals if they were given the opportunity to interact
Study 3
Method
Participants. Forty-nine participants (35 women, 14 men; 39
White, nine Asian, ine Black), with ages ranging from 20 to 62 (M ⫽
35.39, SD ⫽ 9.88), were recruited from a national pool of people
interested in participating in online studies. Participants accessed
the study materials by following a link provided in an e-mail
FLYNN, REAGANS, AND GUILLORY
864
Figure 3.
The positive relationship between need for closure and presumed racial proximity.
advertisement. Upon reaching the study site, participants read a
consent form, completed a personality inventory, performed several cognitive tasks, provided demographic information, and, minutes after the study was completed, received a gift certificate to an
online retailer ($10) in return for their participation.
Procedure. Participants were instructed to complete the Need
for Closure Scale developed by Webster and Kruglanski (1994).
The order in which questions appeared, the scales used to capture
participants’ agreement with each item, and the scoring procedure
were the same as that described in the two previous studies (␣ ⫽
.88). After completing the Need for Closure Scale, participants
moved to the next section of the survey, where they were asked to
read the following instructions:
On the next screen, you will be asked to envision the following scene.
Imagine that you’ve walked into a cafeteria and you see six tables full
of people. Each cafeteria table holds three people on each bench,
which means that you will see a total of 36 people. You will have 30
seconds to review all the photos that are shown. Try to recall where
each of the people shown are located (i.e., where they are sitting and
whom they are sitting next to). After 30 seconds, the picture will
disappear and you will automatically move to the next screen.
After reading the instructions, participants viewed the scene
described above. The 36 photos shown on the screen, including the
16 photos used in Study 2, were head shots of 20 White adults (10
men, 10 women), eight Asian American adults (four men, four
women), and eight African American adults (four men, four
women). Each individual was pictured from the shoulders up.
Following the same steps described in Study 2, a pretest confirmed
that the individuals pictured in these photos were easily recognizable as being White, Asian American, or African American. The
photos were randomly assigned to different “seats” at each of the
cafeteria tables. Each participant saw a different arrangement of
photos in order to minimize the chance that one arrangement
would lead to spurious effects.
After 30 s, the image disappeared, and the participants were
asked to spend the next 10 min working on a set of complex logic
problems. Participants were encouraged to solve as many of the
problems as they possibly could during the time allotted, and they
were prohibited from moving to the next part of the survey until
the entire 10-min period had elapsed. After 10 min, the image of
the cafeteria tables reappeared on the screen, but in this case, the
seats were empty, and the 36 photos appeared at the right side of
the screen in a randomly assigned order. Participants were instructed to “recreate the seating layout shown previously. ‘Drag’
and ‘drop’ the pictures from the pool below to the seating area.”
Participants were allowed to make changes to their selections at
any time. That is, they could move a picture to a designated spot
and then relocate that picture to another spot if they wished to do
so. Participants were not allowed to continue the survey until they
had populated the seating area with all 36 photos. Once the
DO YOU TWO KNOW EACH OTHER?
participant had finished recreating the “cafeteria scene,” they were
allowed to complete the remainder of the survey.
Results
We are interested in the tendency for individuals with a high
NFC to assume that racially similar people sat closer together. We
tested this prediction by calculating the distance between every
photo pair, once using the pair’s position at the original table and
once again using the pair’s position at the participant’s reconstructed table. Our dependent variable (“closeness at the cafeteria
table”) is the ratio of these two distances (the natural log of actual
divided by perceived distance). We created a same-race indicator
variable that was set equal to 1 if members of a photo pair
belonged to the same racial group (i.e., both White, both Asian
American, or both African American) and set equal to 0 otherwise.
We created a separate interaction term (we multiplied the samerace variable and the mean-centered NFC variable) to examine the
extent to which an increase in NFC was associated with a tendency
to place racially similar people closer together.
Because photo pairs were the unit of analysis, we had multiple
observations for each individual respondent (as in Studies 1 and 2).
Once again, we address this concern by estimating a two-way
random effects model. The random intercepts for individual respondents allow us to adjust our standard errors for clustering, and
the random intercepts for photo pairs allow us to control for
unmeasured features of a specific pair that could have affected the
perceived proximity of a photo pair.
The results from the regression analysis are reported in Table 1.
Of greatest interest is the influence of the interaction term on the
dependent variable. As predicted, the coefficient for the interaction
between the same race variable and the NFC variable is positive
and significant (see Table 1, Column 6; ␤ ⫽ .0006, p ⬍ .05). This
indicates that as an individual’s need for closure increases, he or
she was more likely to place racially similar photos closer together.
Next, we distinguished the link between need for closure and
perceived physical proximity in judging different kinds of racial
similarity (White, African American, and Asian American) by
replacing the general same-race variable with three specific samerace variables. The African American indicator variable was set
equal to 1 if both target individuals were African American (0 otherwise). The Asian American indicator variable was set equal to 1 if
both target individuals were Asian American (0 otherwise). Finally,
the White indicator variable was set equal to 1 if both target
individuals were White (0 otherwise). Each of the three specific
same-race indicator variables was multiplied by the mean-centered
NFC variable and included in a separate regression equation. As
shown in Table 1 (Column 7), individuals with a high NFC were
more inclined than those with a low NFC to place White pairings
(␤ ⫽ .0005, p ⬍ .10), African American pairings (␤ ⫽ .0019, p ⬍
.05), and Asian American pairings (␤ ⫽ .0018, p ⬍ .10) closer
together (note the effects for White and Asian American pairings
only approached significance).
Our basic findings are illustrated in Figure 4. The vertical axis
is the proximity of a photo pair at the cafeteria table, and the
horizontal axis is the respondent’s NFC. Two lines are illustrated,
one for when the individuals in the network pair were racially
similar and another for when they were racially dissimilar. The
results indicate that as an individual’s NFC increases, he or she
865
was more likely to place racially similar photos closer together.
Thus, the results support the prediction that high-NFC individuals
were more likely than low-NFC individuals to assume racial
homophily, when racial homophily is defined in terms of physical
proximity.
Again, we considered whether Decisiveness or the remainder of
the NFC scale would have a relatively larger effect on the present
findings by conducting separate analyses for decisiveness (␣ ⫽
.85) and the remainder of the scale (␣ ⫽ .91). The results reveal no
significant impact of decisiveness, in terms of moderating the
effect of racial similarity on proximity (␤ ⫽ .002, p ⫽ .781).
However, as reported in Study 2, the rest of the scale, when used
in place of the overall NFC measure in the interaction term (NFC
multiplied by the general same-race variable), remains a marginally significant predictor of proximity (␤ ⫽ .019, p ⬍ .10). When
decisiveness was removed from the scale, the racial similarity
effect was significant for African Americans (␤ ⫽ .105, p ⬍ .05)
and approached significance for Asians (␤ ⫽ .070, p ⬍ .10), but
was no longer significant for White pairs (␤ ⫽ .011, p ⫽ .34).
Unlike Study 2, these results do not clearly suggest that decisiveness or the remainder of the NFC scale was a relatively stronger
driver of the presumed homophily effect.
General Discussion
On one hand, knowledge about others’ social relations can be a
useful source of social influence (Flynn et al., 2006). On the other
hand, false assumptions about whether two individuals know each
other can be a painful source of social embarrassment. Most of us
can recall an awkward encounter in which two strangers were
casually introduced as though they were friends (e.g., “You two
know each other, right?”), but in fact were not (e.g., “No, we’ve
never met.”). These encounters are especially awkward when the
individuals being introduced are members of the same racial
category. Although people tend to be homophilous in their social
relations, are we likely to assume racial homophily exists? In
particular, are individuals with a high NFC likely to make such
assumptions because they have an appetite for applying social
rules in order to understand social relations?
In the present research, we argue that the tendency to seek order
in judging social networks is more characteristic of some individuals than it is of others. People with a high NFC may have a
tendency to see social networks as having a high degree of closure
by invoking the principle of transitivity as well as a low degree of
racial diversity by invoking the principle of homophily. We found
evidence of the transitivity bias in a set of real relationships (Study
1). In reporting their cognitive networks (i.e., who knows whom?),
high-NFC individuals were more prone to erroneously assume that
two individuals in their social network knew each other. In general, people tend to misperceive social ties by assuming a relationship exists when it does not rather than assuming a relationship
does not exist when in fact it does (e.g., De Soto et al., 1968;
Freeman, 1992; Picek et al., 1975). But this tendency to apply the
rule of transitivity in judging social relations appears to be stronger
for high-NFC individuals than their low-NFC peers.
Beyond projecting network closure, individuals high in NFC
may hold strong assumptions about the affiliation of racially
similar dyads. Specifically, high-NFC individuals are more likely
to assume that two African Americans, two Whites, or two Asian
FLYNN, REAGANS, AND GUILLORY
866
Figure 4.
The positive relationship between need for closure and presumed racial homophily.
Americans are friends, rather than an Asian American and an
African American, or a White person and an Asian American. In
Study 2, when asked to draw network diagrams, high-NFC participants were more likely to draw connections between racially
homophilous pairs. Of course, such reliance on the principle of
homophily in perceiving social relations may be well founded
rather than misguided. Noting this, we examined a recall task in
Study 3, in which participants were asked to recreate a scene with
randomly arranged photos. In performing this task, high-NFC
individuals were more likely to place the racially homophilous
photos closer together, which indicated that high-NFC participants
were more likely than low-NFC participants to exaggerate racial
homophily.
Motives That Might Explain the Link Between Need
for Closure, Transitivity, and Homophily
A recent article by Fu et al. (2007) proposed three powerful
motives for the effects of NFC. First, an individual may be driven
by consensual validation, a concern that drives cultural conformity
according to their findings. Second, those individuals with a high
NFC may be motivated primarily by expediency; that is, they
attempt to minimize the amount of effort expended in making
everyday decisions. Third, according to recent evidence, those
with a high NFC may demonstrate an affinity for political conservatism and the status quo.
In explaining the present findings, the third option seems less
likely, given that (a) the context in which we tested the link
between NFC and transitivity had no political content and (b) the
results supporting the link between NFC and presumed racial
homophily held firm when controlling for attitudes toward diversity, which are often correlated with political conservatism. As for
the first and second motivational accounts, both seem plausible in
helping to explain the present findings. For example, the use of a
heuristic might be efficient, and therefore appealing, for high-NFC
individuals to use in judging complex social relations (effort
minimization). High-NFC individuals evince higher levels of heuristic information processing, including primacy effects (Webster
& Kruglanski, 1994) and the activation of stereotypes (e.g., de
Dreu et al., 1999). Their reliance on heuristics may lead high-NFC
individuals to assume that two of their direct contacts will likely
know each other (because they quickly seize on the transitivity
heuristic) and that members of the same racial category know each
other (because they quickly seize on the homophily heuristic).
Individuals with a high NFC might also perceive higher levels
of transitivity and racial homophily in social relations because it
conforms to their preferences for a shared reality (consensus).
Previous research suggests that high-NFC individuals are more
motivated to reach agreement with their peers (Kruglanski, Webster, & Klem, 1993), are more disturbed by violations of social
norms (Pierro, Mannetti, Kruglanski, & Sleeth-Keppler, 2004),
DO YOU TWO KNOW EACH OTHER?
and have a tendency to perceive collections of individuals as
entitative, homogeneous groups (Kruglanski et al., 2002). Their
taste for consensus might drive high-NFC individuals to assume
the presence of both transitivity and racial homophily in the
broader social system because these features suggest that their
knowledge of social networks is relatively consistent and stable
across situations (Kruglanski & Webster, 1996).
Future Directions
We focus on how NFC can impact an individual’s cognitive
network—their mental representation of social relations. Future
research might build on these findings by examining the link
between an individual’s NFC and his or her actual network relations. Are people with a strong need for cognitive closure bothered
by so-called structural holes—maintaining friendships with two
individuals who are not friends with each other? Do these highNFC individuals feel compelled to introduce their friends to one
another, or connect them in some other way? Maintaining closure
in their personal networks might entail trade-offs for high-NFC
individuals. On one hand, network closure may satisfy their need
for firm social reality. On the other hand, given the strong evidence
that structural holes in one’s social network can yield instrumental
rewards (Burt, 1992), high-NFC individuals might pay a dear price
for maintaining such closure in their social networks.
Future research might also examine different kinds of transitivity. In Study 1, when a high-NFC individual was disconnected
from two colleagues, he or she was more likely to accurately report
those colleagues were also disconnected (“vacuous” transitivity).
When a high-NFC individual was connected to both colleagues, he
or she was more likely to erroneously report the colleagues were
connected. The outcomes are related but could result from different underlying mechanisms. A more in-depth analysis of these
kinds of triads might shed some light on the mechanism underlying
the link between NFC and transitivity. Furthermore, we operationalize NFC as a chronic individual characteristic, but it can also be
situationally induced. Future research might attempt to replicate
the present findings by inducing NFC in a way that highlights a
specific psychological motive. For example, if the effect were to
replicate when using an increase in time pressure, it might indicate
that the “effort minimization” motive was partly responsible for
the outcome.
Another variable that may be worth exploring further is tie
strength or the content of a network connection. An interesting
finding from Studies 2 and 3 was high-NFC individuals’ tendency
to not only link racially similar people but also link them more
closely together, in terms of physical space. Are high-NFC individuals more inclined to not only see connections among other
individuals in their social world but also presume that these connections are strong and meaningful, particularly if the individuals
in question are racially similar? Put differently, individuals with a
high NFC might not stop at assuming that people in their network,
particularly racially similar people, know one another, but that
they know one another well (i.e., they engage in frequent communication and share a familiar rapport).
Following other researchers (e.g., Kruglanski & Webster, 1996;
Shah et al., 1998), we argue that high-NFC individuals are driven
by a desire for “firm social reality”—a sense of permanence and
stability in the social world. This suggests that a potential bound-
867
ary condition for our findings should be the extent to which the
circumstances provide a firm social reality. Our analysis suggests
that NFC will augment the tendency to assume relationships exist
among racially homophilous pairs. But such effects may be attenuated when the context does not activate assumptions about homophily. For example, one might conduct an alternative version of
Study 3, in this case clarifying that individuals seated at the
cafeteria tables were randomly seated. This might undermine highNFC individuals’ motivation to apply a social rule (i.e., homophily) when attempting to recall where these individuals were seated.
As for transitivity, one could conduct the same task as that performed in Study 1, but in this case use inanimate objects instead of
people. If concern for shared reality drives the NFC–transitivity
link, then the same effect should not hold when the targets are not
human beings.
The link between need for cognitive closure and a preference for
similarity may be stronger for some social categories than for
others. For example, we found consistent effects in Studies 2 and
3 for how high-NFC individuals view racially similar and dissimilar dyads. However, in a set of separate analyses, we found no
consistent patterns in how they view same-sex or mixed-sex dyads
in any of the three studies. This suggests that high-NFC individuals
may be more prone to apply a rule of racial homophily rather than
gender homophily in encoding and recalling social relations. Perhaps racial homophily is more in line with high-NFC individuals’
sense of firm social reality (i.e., they may assume that gender
intermixing is much more likely than racial intermixing). Indeed,
there is evidence that individuals both experience more gender
intermixing and observe more gender intermixing (McPherson et
al., 2001). Perhaps high-NFC people would be more motivated to
apply a rule of gender homophily in judging social relations in a
racially homophilous group. This question might be worth exploring in future research.
Conclusion
Previous research has focused on how the NFC can affect
individual cognition. In the present research, we consider whether
such a need can affect the perception of social relations, specifically whether NFC leads to a presumption of network closure or
physical closeness. The evidence suggests that people who have a
high NFC were more likely to assume that other individuals,
particularly those who are racially similar, knew each other or
were close in some way, even when this was not the case. Why
does this happen? It may be that the NFC prompts people to value
“social rules,” such as the principles of homophily and transitivity,
and apply these social rules to their own social context. Our
findings suggest that the application of such rules may be exaggerated, in ways that suggest greater affiliation (same-race or
otherwise) than actually exists.
References
Brass, D. J., & Burkhardt, M. E. (1993). Potential power and power use:
An investigation of structure and behavior. Academy of Management
Journal, 36, 441– 470. doi:10.2307/256588
Burt, R. S. (1992). Structural holes. Cambridge, MA: Harvard University
Press.
Buss, D. M., & Barnes, M. (1986). Preferences in human mate selection.
868
FLYNN, REAGANS, AND GUILLORY
Journal of Personality and Social Psychology, 50, 559 –570. doi:
10.1037/0022-3514.50.3.559
Chirumbolo, A., Livi, S., Mannetti, L., Pierro, A., & Kruglanski, A. W.
(2004). Effects of need for closure on creativity in small group interactions. European Journal of Personality, 18, 265–278. doi:10.1002/
per.518
Coleman, J. S. (1988). Social capital in the creation of human capital.
American Journal of Sociology, 94, S95–S120. doi:10.1086/228943
Coleman, J. S. (1990). Foundations of social theory. Cambridge, MA: The
Belknap Press of Harvard University Press.
de Dreu, C. K., Koole, S. L., & Oldersma, F. L. (1999). On the seizing and
freezing of negotiator inferences: Need for cognitive closure moderates
the use of heuristics in negotiation. Personality and Social Psychology
Bulletin, 25, 348.
De Soto, C. B. (1960). Learning a social structure. Journal of Abnormal
and Social Psychology, 60, 417. doi:10.1037/h0047511
De Soto, C. B., Henley, N. M., & London, M. (1968). Balance and the
grouping schema. Journal of Personality and Social Psychology, 8, 1–7.
doi:10.1037/h0025321
Eberhardt, J. L., Davies, P. G., Purdie-Vaughns, V. J., & Johnson, S. L.
(2006). Looking deathworthy. Psychological Science, 17, 383–386. doi:
10.1111/j.1467-9280.2006.01716.x
Flynn, F. J., Reagans, R. E., Amanatullah, E. T., & Ames, D. R. (2006).
Helping one’s way to the top: Self-monitors achieve status by helping
others and knowing who helps whom. Journal of Personality and Social
Psychology, 91, 1123. doi:10.1037/0022-3514.91.6.1123
Freeman, L. C. (1992). Filling in the blanks: A theory of cognitive
categories and the structure of social affiliation. Social Psychology
Quarterly, 55, 118 –127. doi:10.2307/2786941
Fu, J. H., Morris, M. W., Lee, S., Chao, M., Chiu, C., & Hong, Y. (2007).
Epistemic motives and cultural conformity: Need for closure, culture,
and context as determinants of conflict judgments. Journal of Personality and Social Psychology, 92, 191–207.
Goldstone, R. L. (2000). Unitization during category learning. Journal of
Experimental Psychology: Human Perception and Performance, 26,
86 –112. doi:10.1037/0096-1523.26.1.86
Granovetter, M. S. (1973). The strength of weak ties. American Journal of
Sociology, 78, 1360. doi:10.1086/225469
Holland, P. W., & Leinhardt, S. (1971). Transitivity in structural models of
small groups. Small Group Research, 2, 107–124.
Ibarra, H., Kilduff, M., & Tsai, W. (2005). Zooming in and out: Connecting
individuals and collectivities at the frontiers of organizational network
research. Organization Science, 16, 359. doi:10.1287/orsc.1050.0129
Janicik, G. A., & Larrick, R. P. (2005). Social network schemas and the
learning of incomplete networks. Journal of Personality and Social
Psychology, 88, 348 –364. doi:10.1037/0022-3514.88.2.348
Jost, J. T., Glaser, J., Kruglanski, A. W., & Sulloway, F. J. (2003). Political
conservatism as motivated social cognition. Psychological Bulletin, 129,
339 –375. doi:10.1037/0033-2909.129.3.339
Kilduff, M., Crossland, C., Tsai, W., & Krackhardt, D. (2008). Organizational network perceptions versus reality: A small world after all?
Organizational Behavior and Human Decision Processes, 107, 15–28.
doi:10.1016/j.obhdp.2007.12.003
Krackhardt, D. (1987). Cognitive social structures. Social Networks, 9,
109 –134. doi:10.1016/0378-8733(87)90009-8
Krackhardt, D., & Kilduff, M. (1999). Whether close or far: Social distance
effects on perceived balance in friendship networks. Journal of Personality and Social Psychology, 76, 770 –782. doi:10.1037/00223514.76.5.770
Kruglanski, A. W. (1990). Motivations for judging and knowing: Implications for causal attribution. In E. T. Higgins & R. M. Sorrentino (Eds.),
The handbook of motivation and cognition (Vol. 2, pp. 333–386). New
York, NY: Guilford Press.
Kruglanski, A. W., Atash, M. N., DeGrada, E., Mannetti, L., Pierro, A., &
Webster, D. M. (1997). Psychological theory testing versus psychometric nay-saying: Comment on Neuberg et al.’s (1997) critique of the Need
for Closure Scale. Journal of Personality and Social Psychology, 73,
1005–1016. doi:10.1037/0022-3514.73.5.1005
Kruglanski, A. W., Shah, J. Y., Pierro, A., & Mannetti, L. (2002). When
similarity breeds content: Need for closure and the allure of homogeneous and self-resembling groups. Journal of Personality and Social
Psychology, 83, 648 – 662. doi:10.1037/0022-3514.83.3.648
Kruglanski, A. W., & Webster, D. M. (1996). Motivated closing of the
mind: “Seizing” and” freezing.” Psychological Review, 103, 263. doi:
10.1037/0033-295X.103.2.263
Kruglanski, A. W., & Webster, D. M., & Klem, A. (1993). Motivated
resistance and openness to persuasion in the presence or absence of prior
information. Journal of Personality and Social Psychology, 65, 861–
876.
Laumann, E. O., & Knoke, D. (1987). The organizational state: Social
choice in national policy domains. Madison: University of Wisconsin
Press.
Lin, N. (2001). Social capital: A theory of social structure and action.
London, England: Cambridge University Press.
Louch, H. (2000). Personal network integration: Transitivity and homophily in strong-tie relations. Social Networks, 22, 45– 64. doi:10.1016/
S0378-8733(00)00015-0
Mannetti, L., Pierro, A., Kruglanski, A., Taris, T., & Bezinovic, P. (2002).
A cross-cultural study of the Need for Cognitive Closure Scale: Comparing its structure in Croatia, Italy, USA and the Netherlands. British
Journal of Social Psychology, 41, 139 –156. doi:10.1348/
014466602165108
McPherson, M., Smith-Lovin, L., & Cook, J. M. (2001). Homophily in
social networks. Annual Review of Sociology, 27, 415– 444.
Neuberg, S. L., Judice, T. N., & West, S. G. (1997). What the Need for
Closure Scale measures and what it does not: Toward differentiating
among related epistemic motives. Journal of Personality and Social
Psychology, 72, 1396 –1412. doi:10.1037/0022-3514.72.6.1396
Newcomb, T. M. (1961). The acquaintance process. New York, NY: Holt,
Rinehart, & Winston.
Picek, J. S., Sherman, S. J., & Shiffrin, R. M. (1975). Cognitive organization and coding of social structures. Journal of Personality and Social
Psychology, 31, 758 –768. doi:10.1037/0022-3514.31.4.758
Pierro, A., Mannetti, L., Kruglanski, A. W., & Sleeth-Keppler, D. (2004).
Relevance override: On the reduced impact of “cues” under highmotivation conditions of persuasion studies. Journal of Personality and
Social Psychology, 86, 251–264.
Podolny, J., & Baron, J. (1997). Resources and relationships: Social
networks and mobility in the workplace. American Sociological Review,
62, 673– 693. doi:10.2307/2657354
Rabe-Hesketh, S., & Skrondal, A. (2005). Multilevel and longitudinal
modeling using Stata. College Station, TX: Stata Corp.
Reagans, R., & Zuckerman, E. (2008). All in the family. Industrial and
Corporate Change, 17, 979 –999. doi:10.1093/icc/dtn037
Rubin, Z., & Zajonc, R. B. (1969). Structural bias and generalization in the
learning of social structures. Journal of Personality, 37, 310 –324. doi:
10.1111/j.1467-6494.1969.tb01747.x
Sacco, J. M., Scheu, C. R., Ryan, A. M., & Schmitt, N. (2003). An
investigation of race and sex similarity effects in interviews: A multilevel approach to relational demography. Journal of Applied Psychology, 88, 852– 865. doi:10.1037/0021-9010.88.5.852
Shah, J. Y., Kruglanski, A. W., & Thompson, E. P. (1998). Membership
has its (epistemic) rewards: Need for closure effects on in-group bias.
Journal of Personality and Social Psychology, 75, 383–393. doi:
10.1037/0022-3514.75.2.383
Tetlock, P. E. (2005). Expert political judgment: How good is it? How can
we know? Princeton, NJ: Princeton University Press.
Van Hiel, A., & Mervielde, I. (2003). The need for closure and the spontane-
DO YOU TWO KNOW EACH OTHER?
ous use of complex and simple cognitive structures. The Journal of Social
Psychology, 143, 559 –568. doi:10.1080/00224540309598463
von Hecker, U. (1997). How do logical inference rules help construct social
mental models? Journal of Experimental Social Psychology, 33, 367–
400. doi:10.1006/jesp.1997.1325
Watson, D., Klohnen, E. C., Casillas, A., Nus, S. E., Haig, J., & Berry,
D. S. (2004). Match makers and deal breakers: Analyses of assortative
mating in newlywed couples. Journal of Personality, 72, 1029 –1068.
doi:10.1111/j.0022-3506.2004.00289.x
Webster, D. M., & Kruglanski, A. W. (1994). Individual differences in
need for cognitive closure. Journal of Personality and Social Psychology, 67, 1049 –1062. doi:10.1037/0022-3514.67.6.1049
Zajonc, R. B., & Burnstein, E. (1965a). The learning of balanced and
869
unbalanced social structures. Journal of Personality, 33, 153–163. doi:
10.1111/j.1467-6494.1965.tb01378.x
Zajonc, R. B., & Burnstein, E. (1965b). Structural balance, reciprocity, and
positivity as sources of cognitive bias. Journal of Personality, 33,
570 –583. doi:10.1111/j.1467-6494.1965.tb01403.x
Zajonc, R. B., & Sherman, S. J. (1967). Structural balance and the induction of relations. Journal of Personality, 35, 635– 650. doi:10.1111/
j.1467-6494.1967.tb01453.x
Received June 4, 2009
Revision received April 26, 2010
Accepted May 10, 2010 䡲
Online First Publication
APA-published journal articles are now available Online First in the PsycARTICLES database.
Electronic versions of journal articles will be accessible prior to the print publication, expediting
access to the latest peer-reviewed research.
All PsycARTICLES institutional customers, individual APA PsycNET威 database package subscribers, and individual journal subscribers may now search these records as an added benefit.
Online First Publication (OFP) records can be released within as little as 30 days of acceptance and
transfer into production, and are marked to indicate the posting status, allowing researchers to
quickly and easily discover the latest literature. OFP articles will be the version of record; the
articles have gone through the full production cycle except for assignment to an issue and
pagination. After a journal issue’s print publication, OFP records will be replaced with the final
published article to reflect the final status and bibliographic information.